Solar Farm Digital Twin

Overview

The LSU Solar Farm project builds a digital twin of the university's solar installation. I joined in 2025 as the rendering engineer, responsible for two systems: a heatmap that visualizes how much sunlight reaches every part of the site, and a plant generator that will grow vegetation according to that light.

Light-intensity heatmap

The heatmap reads the sunlight information in the scene and replaces it with colour by intensity: fully lit areas turn red, shadowed areas blue, and everything in between blends smoothly — turning the whole level into a readable map of exposure.

It's a compute shader living inside a plugin. Rather than modify the engine, it hooks into the renderer through a custom Scene View Extension, with an Engine Subsystem managing its lifetime, so it can be dropped into any project.

The view-dependence problem

The first version derives intensity from the GBuffer. That's fast, but the GBuffer only describes what the current camera sees — so the heatmap changes as the camera moves, which is wrong for an analysis tool. I'm now working on a version that estimates light intensity with basic ray tracing, making the result camera-independent and more accurate.

Light-driven plant growth

The second system uses Unreal's PCG framework with its GPU compute path. So far I've used it to generate fractal structures directly on the GPU; the next step connects it to the light-intensity data so plants spawn and grow — or shrink back — according to how much sun each area receives.

Credit to renderBucket's PCG Compute introduction for an excellent explanation of the framework.